Skip to main content

Use LLM embeddings to create mappings between the Brick ontology and external vocabularies

Project description

Brick-vocab-mapper

pip install brickmapper

This package is a simple wrapper around the OpenAI embeddings API to assist in producing vocabulary mappings between external vocabs (like Haystack protos, gbXML equipment names, or internal naming schems) and Brick classes.

You will need an OpenAI API key to compute your own embeddings. This costs money, but is very cheap (much less than $1) to do for hundreds of terms. We cache embeddings in .index files to avoid hitting the API again. The brick.index file in this repo will allow you to run the example below without any OpenAI API calls.

You must provide your external vocabulary to the Brick mapper tool as a list of dictonaries. Each dictionary defines key-value pairs of a single concept; this dictionary must contain a key called name which contains the unique name of the term to be mapped.

There are two kinds of mappings that can be produced by the package:

  • stable: a stable mapping performs a 1-1 matching of external terms to Brick classes; it does not permit overlaps
  • unstable: an unstable mapping allows 1 external term to match multiple Brick classes and vice-versa

Haystack Protos Example

from brickmapper import Mapper
from rdflib import BRICK
import urllib.request
import csv

defns = []
# read haystack proto definitions off their website
request = urllib.request.Request("https://project-haystack.org/download/protos.csv", headers={'User-Agent': "Python script"})
response = urllib.request.urlopen(request)
rdr = csv.DictReader(response.read().decode('utf8').splitlines())
for row in rdr:
    # save each proto (e.g. 'air temp sensor point') as a dictionary
    defns.append({'name': row['proto']})

# create a new Mapper object which knows about the definitions. This will
# access OpenAI to compute embeddings if necessary
m = Mapper(defns, "haystack_proto.index")
# define which set of Brick classes we want to match against
# all points which aren't parameters (these have similar enough names it can confuse the mapping)
points = m.get_brick_classes(BRICK.Point, [BRICK.Parameter])
# all equipment in Brick
equips = m.get_brick_classes(BRICK.Equipment)
# concatenate these two lists of Brick classes
bcs = points + equips

# compute a 'stable' mapping (1:1 mapping)
stable = m.get_mapping(bcs, allow_collisions=False)
# compute an 'unstable' mapping that allows 1:n and n:1 matches
unstable = m.get_mapping(bcs, allow_collisions=True)

# loop through and print the stable/unstable matches
for k,v in unstable.items():
    print(f"{k}:\n\tunstable: {v}\n\tstable: {stable.get(k)}")

Gives output like:

flue hot water heat valve cmd point:
        unstable: https://brickschema.org/schema/Brick#Domestic_Hot_Water_Valve
        stable: https://brickschema.org/schema/Brick#Hot_Water_Valve
steam bypass pump point:
        unstable: https://brickschema.org/schema/Brick#Bypass_Valve
        stable: https://brickschema.org/schema/Brick#Bypass_Valve
makeup water entering valve actuator equip:
        unstable: https://brickschema.org/schema/Brick#Makeup_Water_Valve
        stable: https://brickschema.org/schema/Brick#Makeup_Water_Valve
makeup water entering pipe equip:
        unstable: https://brickschema.org/schema/Brick#Eye_Wash_Station
        stable: https://brickschema.org/schema/Brick#Eye_Wash_Station
steam leaving pump point:
        unstable: https://brickschema.org/schema/Brick#Emergency_Wash_Station
        stable: https://brickschema.org/schema/Brick#Steam_Baseboard_Radiator
infiltration well equip:
        unstable: https://brickschema.org/schema/Brick#Intrusion_Detection_Equipment
        stable: https://brickschema.org/schema/Brick#Intrusion_Detection_Equipment
economizer damper actuator equip:
        unstable: https://brickschema.org/schema/Brick#Economizer_Damper
        stable: https://brickschema.org/schema/Brick#Economizer_Damper
steam leaving point:
        unstable: https://brickschema.org/schema/Brick#Steam_Distribution
        stable: https://brickschema.org/schema/Brick#Steam_Distribution
flue equip:
        unstable: https://brickschema.org/schema/Brick#Fire_Safety_Equipment
        stable: https://brickschema.org/schema/Brick#Fire_Safety_Equipment
condenser water entering pump equip:
        unstable: https://brickschema.org/schema/Brick#Condenser_Water_Pump
        stable: https://brickschema.org/schema/Brick#Condenser_Water_Pump

TODOs

  • incorporate common abbreviations & support embedding of them

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

brickmapper-0.1.3.tar.gz (6.0 kB view details)

Uploaded Source

Built Distribution

brickmapper-0.1.3-py3-none-any.whl (6.4 kB view details)

Uploaded Python 3

File details

Details for the file brickmapper-0.1.3.tar.gz.

File metadata

  • Download URL: brickmapper-0.1.3.tar.gz
  • Upload date:
  • Size: 6.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.7.1 CPython/3.12.1 Darwin/23.2.0

File hashes

Hashes for brickmapper-0.1.3.tar.gz
Algorithm Hash digest
SHA256 11618e3a9bed66dd5f94f09d7cb406469edb62eb505344c90181c254b0e0c5d6
MD5 222964e81171f4f4aed313ef566fd220
BLAKE2b-256 6da657628a34bec81c6aeb2c364faad584b9d93c706def7fa1387a0b26935288

See more details on using hashes here.

File details

Details for the file brickmapper-0.1.3-py3-none-any.whl.

File metadata

  • Download URL: brickmapper-0.1.3-py3-none-any.whl
  • Upload date:
  • Size: 6.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.7.1 CPython/3.12.1 Darwin/23.2.0

File hashes

Hashes for brickmapper-0.1.3-py3-none-any.whl
Algorithm Hash digest
SHA256 155a8322771bef966d9d7f6616ac1f517311e34eb4385ee362e923073dfd581d
MD5 7eb0887fe5f7e737f9ebb2a508f479df
BLAKE2b-256 659ccd233a5294116ade15d4f0c25178afad22d7061e6c2bddddccf611610508

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page